The vast amount of healthcare data to diagnose diseases, progression tracking over time and real-time patient monitoring make it essential that we can accurately analyze such dynamic signals for better treatments. Machine learning models extract information from such complex data and create predictive power for cardiovascular diseases using their models. However, recent studies have indicated that these models are vulnerable when adversarial attacks result in wrong predictions. In this work, we integrated adversarial learning into the prediction models of cardiovascular disease to enhance robustness. We have compared the performance of different models, such as SVMs, random forests, and sequential models. Among them, the Sequential model was the best, with an accuracy of 97%. In this model, we applied the Fast Gradient Sign Method to make it robust against adversarial perturbations. Results show that this significantly enhances model robustness with no compromise in performance and thus is reliable for clinical settings.